Update app.py

app.py

@@ -3,7 +3,7 @@ import shap
 import numpy as np
 import scipy as sp
 import torch
-import tensorflow as tf
+# import tensorflow as tf  <-- Removed to match your requirements
 import transformers
 from transformers import pipeline
 from transformers import RobertaTokenizer, RobertaModel
@@ -15,7 +15,6 @@ import csv
 import os
 
 HF_TOKEN = os.getenv("hf_token")
-
 csv.field_size_limit(sys.maxsize)
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
 
@@ -32,31 +31,26 @@ explainer = shap.Explainer(pred)
 # NER pipeline
 ner_tokenizer = AutoTokenizer.from_pretrained("d4data/biomedical-ner-all")
 ner_model = AutoModelForTokenClassification.from_pretrained("d4data/biomedical-ner-all")
-ner_pipe = pipeline("ner", model=ner_model, tokenizer=ner_tokenizer, aggregation_strategy="simple") # pass device=0 if using gpu
+ner_pipe = pipeline("ner", model=ner_model, tokenizer=ner_tokenizer, aggregation_strategy="simple")
 
-# def adr_predict(x):
 def adr_predict(x):
     # Ensure input is treated as a string
     text_input = str(x).lower()
-
-    encoded_input = tokenizer(text_input, return_tensors='pt').to(device) # Move input to device
+    encoded_input = tokenizer(text_input, return_tensors='pt').to(device)
     output = model(**encoded_input)
     
-    scores = torch.softmax(output.logits, dim=-1)[0].detach().cpu().numpy() # Apply softmax on logits, move to cpu and convert to numpy
-
+    scores = torch.softmax(output.logits, dim=-1)[0].detach().cpu().numpy()
+    
     try:
         shap_values = explainer([text_input])
-
-        local_plot = shap.plots.text(shap_values[0], display=False) 
-
+        local_plot = shap.plots.text(shap_values[0], display=False)
     except Exception as e:
         print(f"SHAP explanation failed: {e}")
-        local_plot = "<p>SHAP explanation not available.</p>" # Provide a fallback
+        local_plot = "<p>SHAP explanation not available.</p>" 
 
     # NER processing
     try:
         res = ner_pipe(text_input)
-
         entity_colors = {
             'Severity': 'red',
             'Sign_symptom': 'green',
@@ -68,29 +62,24 @@ def adr_predict(x):
         }
         htext = ""
         prev_end = 0
-
         res = sorted(res, key=lambda x: x['start'])
-
         for entity in res:
             start = entity['start']
             end = entity['end']
-            word = text_input[start:end] # Extract original text segment
+            word = text_input[start:end]
             entity_type = entity['entity_group']
-            color = entity_colors.get(entity_type, 'lightgray') # Use get with a default color
-
-            # Append text before the entity
+            color = entity_colors.get(entity_type, 'lightgray')
+            
             htext += f"{text_input[prev_end:start]}"
-            # Append the highlighted entity
             htext += f"<mark style='background-color:{color};'>{word}</mark>"
             prev_end = end
-        # Append any remaining text after the last entity
+            
         htext += text_input[prev_end:]
     except Exception as e:
         print(f"NER processing failed: {e}")
-        htext = "<p>NER processing not available.</p>" # Provide a fallback
+        htext = "<p>NER processing not available.</p>"
 
     label_output = {"Severe Reaction": float(scores[1]), "Non-severe Reaction": float(scores[0])}
-
     return label_output, local_plot, htext
 
 def main(prob1):
@@ -99,33 +88,26 @@ def main(prob1):
 title = "Welcome to **ADR Detector** 🪐"
 description1 = """This app takes text (up to a few sentences) and predicts to what extent the text describes severe (or non-severe) adverse reaction to medicaitons. Please do NOT use for medical diagnosis."""
 
-# Use the 'with' syntax for Blocks
 with gr.Blocks(title=title) as demo:
     gr.Markdown(f"## {title}")
     gr.Markdown(description1)
     gr.Markdown("""---""")
-
-    # Define input and output components
-    prob1 = gr.Textbox(label="Enter Your Text Here:",lines=2, placeholder="Type it here ...")
-
-    # Output components matching the return values of the main function
-    label = gr.Label(label = "Predicted Label")
-    local_plot = gr.HTML(label = 'Shap Explanation') 
+    
+    prob1 = gr.Textbox(label="Enter Your Text Here:", lines=2, placeholder="Type it here ...")
+    
+    label = gr.Label(label="Predicted Label")
+    local_plot = gr.HTML(label='Shap Explanation') 
     htext = gr.HTML(label="Named Entity Recognition") 
-
+    
     submit_btn = gr.Button("Analyze")
-
     submit_btn.click(
-        fn=main, # The function to call
-        inputs=[prob1], # The input components
-        outputs=[label, local_plot, htext], # The output components
-        api_name="adr" # Keep the api_name if you intend to use the API
+        fn=main,
+        inputs=[prob1],
+        outputs=[label, local_plot, htext],
+        api_name="adr"
     )
-
-    # Examples section
+    
     gr.Markdown("### Click on any of the examples below to see how it works:")
-    # Gradio 4.0+ Examples usage. Pass inputs and outputs components directly.
-    # cache_examples is recommended for faster loading of examples.
     gr.Examples(
         examples=[
             ["A 35 year-old male had severe headache after taking Aspirin. The lab results were normal."],
@@ -133,10 +115,9 @@ with gr.Blocks(title=title) as demo:
         ],
         inputs=[prob1],
         outputs=[label, local_plot, htext],
-        fn=main, # Provide the function to run for caching examples
+        fn=main,
         cache_examples=False,
         run_on_click=True
     )
 
-# Launch the demo
 demo.launch()